Extrapolation of dose-response carcinogen data from animals to establish "safe" levels for human exposures (1 in a million) is challenging as it requires modeling of dose-response over at least 4 orders of magnitude. In many cases, the conservative approach is assumed linearity. The largest tumor study in a rodent model used 24,192 mice to detect 1 cancer in 100. The trout tumor model has proven valuable in identification of carcinogens and their mechanism of action. Trout are very sensitive to the hepatocarcinogenicity of aflatoxin B1 (AFB1), the only IARC human carcinogen where exposure is through the food supply. Hepatocellular carcinoma (HCC) is responsible for over 600,000 deaths a year worldwide and accounts for 10-15% of all deaths in certain regions. HCC is the most rapidly increasing solid tumor in the U.S. For AFB1, the target organ, metabolism, DNA adduction and gene targets are similar in trout and human and, in this example, the trout model is superior to mouse. The trout model can utilize large numbers of animals to evaluate cancer across a wide range of doses. This approach is possible due to a number of advantages of this model including low spontaneous tumor incidence and low per diem costs. We have recently completed the largest cancer study in any animal model, utilizing 42,000 trout to assess carcinogenicity of dibenzo[a,l]pyrene (DBP) at ultra-low doses. The target was an order of magnitude lower than the mouse ED01 study, to one cancer in 1000. Our data established a dose of DBP that resulted in 1 cancer in 5,000 trout and the dose-response was non-linear. We now propose to utilize this ultra-low dose model to test the hypothesis that the non-linearity of cancer incidence at ultra-low dose also applies to AFB1. Tumor incidence data will be coupled with measurements of molecular dosimetry, cell proliferation, apoptosis and gene expression utilizing a custom microarray in order to test the second hypothesis, that in contrast to tumor incidence, these biomarkers of carcinogenicity exhibit linearity across the entire tumor dose-response range.